On Mon, 18 Mar 2002 06:16:39,Valdis Kletnieks wrote : > On Mon, 18 Mar 2002 09:11:14 +0300, Oleg Drokin said: > > > It is somehow related to knfsd, I think. I will tell you what I will be > > able to find. > > > The only thing that ext{2,3}/reiserfs differs right now is buffer flushing > > policy, while ext{2,3} just marks block dirty and returns, reiserfs > > actually waits for disk to finish it's I/O. > > If that *doesnt* explain the performance differences, I'd be surprised....
That could be well explain the "little" stalls I see during all latest dbench test, here. Latest Kernel is: 2.4.19-pre3-dn1 (;-) AA vm_29 O(1) preemption+lock-break Under heavy IO (dbench) there are times in which there is no progress and then fast drawing of the little dots, again. Regards, Dieter -- Dieter Nützel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: [EMAIL PROTECTED]